A few top OpenAI researchers considered joining Ilya Sutskever’s new company, SSI(Safe SuperIntelligence), but OpenAI countered with $2 million retention bonuses and equity increases worth up to $20 million to keep them.
- Autonomous software engineers from OpenAI are coming. Sarah Friar(CFO @OpenAI) yesterday "And then the third(agent) that is coming is what we call 'A-SWE'. The agentic software engineer. And this is not just augmenting the current software engineers in your workforce,
- Read Noam's thread carefully. Winning a gold medal at the 2025 IMO is an outstanding achievement, but in some ways, it might just be noise that grabbed the headlines. They have recently developed new techniques that work much better on hard-to-verify problems, have extended TTCToday, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as humans, without tools. As remarkable as that sounds, it’s even more significant than the headline 🧵
- Sam Altman(@sama) at the Federal Reserve Capital Framework conference in Washington: "I'm very interested in what it would mean to give everyone on Earth free copies of GPT-5, running for them all the time, with every business truly enabled by this level of technology."
- New Ilya Sutskever (@ilyasut) video dropped yesterday. <Dr. Ilya Sutskever, Honorary Doctor of Philosophy Degree for the Year 2025> "...the idea that an AI can build the next generation of AI. The intellgence explosion. Holy moly, what do you do about that?"
00:00 - Alec Radford and Bob McGrew have joined Mira Murati's Thinking Machine Lab as advisors.
- Scores for o3 and o4-mini have been uploaded for USAMO 2025. Gemini 2.5 pro: 24.40% / $6.23 o3 (high): 21.73% / $24.17 o4-mini (high): 19.05% / $2.21
- Insights from the OpenAI IMO Gold Medal Team Podcast (@alexwei_ , @polynoamial, @SherylHsu02) 1. Their dedicated effort towards this year’s IMO lasted only about two months, which was a very short preparation period. 2. The core team consisted of just three members. (Of
- Sam Altman(@sama) acknowledges that hallucinations increased in the transition from o1 to o3. However, he says that things will be much better in the next version, and that they have learned a lot about how to align reasoning models. He suspects that people will be very happy
- 1. (11:03~) Theo Von: Yeah, you'll sneeze and release 5.0, dude. You'll freaking... Are you guys at 4.5 already? Sam Altman: We're 4.5 already. 5.0 is uh... I think it's going to be great. 2. (42:05~) Sam Altman: GPT-5 is the smartest thing. GPT-5 is smarter than us in almostSam Altman: "This morning I was testing our new model and I got a question. I got emailed a question that I didn't quite understand. And I put it in the model, this GPT-5, and it answered it perfectly. And I really kind of sat back in my chair and I was just like, oh man,
- Dwarkesh(@dwarkesh_sp) mentioned that he heard rumors suggesting Ilya's SSI might be working on something related to TTT (Test-Time Training). He speculated that it could be connected to continual learning.
- Unfortunately, OpenAI's open-source models performed terribly on my vibe test. Here are the conclusions I drew from looking at the output: 1. Scores similar to Qwen-3 coder suggest a heavy optimization for STEM tasks. 2. The multilingual capabilities (in this case, Korean) are
- It seems my concerns were valid. This is the result of re-running the tests after changing the provider setting from the default (which automatically routed to Groq) to Fireworks. To emphasize again, the only thing I changed was explicitly fixing the provider in the code. All
- Story from the OpenAI Agent Team (@isafulf, @caseychu9, @EdwardSun0909) 1. The newly released ChatGPT agent was created through a collaboration between the Operator team and the Deep Research team, which were very small groups of around 4 and 6 members respectively. 2. The
